Coronary vessel trees from 3D imagery: a topological approach

Area of Science:

• Medical Imaging
• Computer Vision
• Biomedical Engineering

Background:

• Vascular tree reconstruction is crucial for analyzing blood flow and diagnosing diseases.
• Existing methods often struggle with low contrast and spurious features in 3D medical images.

Purpose of the Study:

• To present a novel, simple algorithm for accurate vascular tree reconstruction from 3D image data.
• To enhance the extraction of coronary artery trees from cardiac CT scans.

Main Methods:

• The algorithm identifies persistent intensity maxima across all axis-aligned 2D slices.
• A graph-based approach connects these maxima to form an initial vascular forest.
• Geometric filters are applied to refine the forest, removing noise and filling gaps.

Main Results:

• The method successfully approximates vascular structures, particularly blood vessels.
• Post-processing filters effectively reduce spurious features and connect fragmented segments.
• Successful extraction of coronary vascular trees from heart CT scans was demonstrated.

Conclusions:

• The proposed method offers a robust and simple approach for vascular tree reconstruction.
• It shows significant potential for clinical applications, especially in cardiac imaging.
• Further refinement could enhance accuracy in complex vascular networks.